[Python-Dev] Better unittest failures